基于All-confidence的正负关联分类

针对分类中如何有效利用负关联模式提高分类准确率，提出了一种基于正负关联模式的分类算法．利用类Apriori算法挖掘包含正项或／和负项且项与项之间互相关联的正负关联模式来产生分类规则．为提高挖掘效率，先找出能覆盖训练集的信息熵最小k个正，负项．然后，把这k个正／负项分别与其他项进行连接得到相应的正负关联模式．实验表明，该算法有效减少了挖掘的规则数，极大减少了挖掘时间，并提高了分类准确率．

All-confidence based Positive and Negative Association Classification

To take advantage of the negative association patterns for improving classification accuracy, a new algorithm called classification based on positive and negative association patterns （APNAC） is proposed. APNAC uses Apriori-like algorithm to mine such itemsets which contain positive or/and negative item and in which items are associated. In order to efficiently mining, items with weak classification power in terns of information entropy are pruning, and the positive and negative association patterns must contain one of best k items. Our experiments show that APNAC efficiently decreases the number of rules and improves the classification accuracy.